If you build with AI agents: is packaging knowledge into reusable skills a real pain, or am I imagining it?
I've been building with AI agents for a while, and the thing that keeps biting me isn't the model — it's that every agent is only as good as the knowledge you hand it. I keep finding myself copying a doc into a chat, reformatting it into steps, adding a "when should you use this," writing a little done-checklist, and then doing that same dance over and over for every new workflow. It's the boring 20 minutes nobody talks about, and it feels like the reason most people's agents stay generic.
Before I build anything heavier around this, I want to sanity-check whether the pain is actually real for other people or just my own workflow being messy.
So, if you build with agents: how are you handling reusable knowledge today — hand-writing skill files, ignoring it entirely, or using some tool I don't know about? How much time does it actually eat? And if something could turn a raw doc or SOP i...